Abstract: DEM data with high-resolution and continuous surface is the main data source for obtaining lunar morphological features and performing digital terrain analysis. A region of Chang'E-5 candidate landing region is selected as the experimental area. First, high-resolution DEM data and corresponding DOM images are generated based on LROC NAC stereo images, ISIS3 and Stereo Pipeline software, and compared with Japanese SELENE (KAGUYA) TC images; Then, Inverse Distance Weighting, Radial Basis Function and Empirical Bayesian Kriging are used to interpolate the hollow areas of DEM data and cross validation analysis was performed for three different interpolation methods. The results show that the resolution of DEM and DOM generated in this study is about 3.5 m, which is obviously clearer than that of 7.4 m resolution Japanese TC images, and has stronger terrain expression ability; Radial Basis Function interpolation is the best method for hollow interpolation, and the mean square error of cross-validation is 0.26 m. The study has a certain role in accurately acquiring lunar morphological features and locating detectors, and provides a reference for high resolution continuous DEM data generation in other regions.

Key words: DEM data, interpolation method, LROC NAC image, DOM image, Chang'E-5 candidate landing region